CAD4TB has been developed by Delft Imaging in cooperation with Radboud Universiteit and the Lung Institute in Cape Town. The involvement of academic researchers from the beginning demonstrates the intention to deliver a high-quality and reliable product.

Click here to discover complete list of scientific publications on CAD4TB

Publication highlights on CAD4TB Performance

• CAD4TB performs significantly better than human readers [Tanzania; Breuninger et al., PLOS One, 2014], [Qin et al., The Lancet Digital Health, 2021]
• CAD4TB on par with human expert readers [South Africa; Hogeweg et al., IEEE Trans Med Imaging, 2015], [Tanzania; Steiner et al., Public Health Action, 2015], [Zambia; Melendez et al., IJTLD, 2017], [London; Melendez et al., IJTLD, 2018], [the Philippines; Philipsen et al., IJTLD, 2019], [Pakistan; Murphy et al., Nature Scientific Reports, 2020] [South Africa; Fehr et al., medRxiv, 2021], [Tavaziva et al., Clinical Infectious Diseases, 2021]
• CAD4TB met the WHO’s TPP at 90% sensitivity [Qin et al., The Lancet Digital Health, 2021]
• CAD4TB could perform similarly to an expert radiologist in triaging TB and facilitate high-throughput TB screening in South Africa, and has potential to replace radiologists for triaging CXRs in prevalence surveys [Fehr et al., medRxiv, 2021]
• CAD4TB performance was significantly better than a clinical officer in Tanzania [Breuninger et al., PLOS One, 2014]

Publication highlights on CAD4TB Efficiency

• Using CAD4TB can save costs on molecular tests such as Xpert tests [Bangladesh; Rahman et al., European Respiratory Journal, 2017], [Pakistan; Zaidi et al., Nature Scientific Reports, 2018], [Pakistan; Habib et al., Scientific Reports, 2020], [Qin et al., The Lancet Digital Health, 2021]
• All five AI algorithms significantly outperformed radiologists and reduced the number of Xpert tests required by 50% while maintaining a sensitivity above 90% [Qin et al., The Lancet Digital Health, 2021]
• CAD4TB as a triage tool to pre-screen people with diabetes for Xpert testing can improve case-detection while potentially reducing program costs with more rational use of expensive molecular tests [Habib et al., Nature Scientific Reports, 2020]
• Cost per screened subject using CAD4TB v6 is almost half in comparison to screening without, while maintaining almost 2.5x higher daily throughput in Pakistan [Murphy et al., Nature Scientific Reports, 2020]
• Used as a pre-screening system in TB-endemic regions, CAD4TB allows for testing of much larger numbers of subjects at a fraction of the cost [Murphy et al., Nature Scientific Reports, 2020]
• CAD4TB as a second reader showed 22% increase of TB detection and could help detect additional TB-positive cases that were initially missed in the Philippines at limited additional costs [Philipsen et al., IJTLD, 2019]
• CAD4TB as a triage tool could minimize use of expensive molecular tests in Pakistani low resource settings [Zaidi et al., Nature Scientific Reports, 2018]
• Reduces cost per identified TB case and the cost per screened subject by almost half [Philipsen et al., Nature Scientific Reports, 2015]
• Useful where trained human readers are scarce [Muyoyeta et al., PLOS One, 2014]

Publication highlights on CAD4TB for High-risk group / HIV, Diabetes & Prison-screening

HIV

• Community-based multi-disease screening effort in HIV-endemic area in rural South Africa indicates that CAD4TB has the potential to replace human readers for CXR triaging in TB prevalence surveys [Fehr et al., medRxiv, 2021]
• Digital CXR with CAD4TB with universal HIV screening significantly increased timelines and completeness of HIV and TB diagnosis in Malawi [Malawi; MacPherson et al., PLOS MEDICINE, 2021]
• Combined use of mobile X-ray and CAD4TB to identify TB among high-risk groups to improve early TB diagnosis can deliver a major contribution to TB control in Romania [Mahler et al., BMJ Open, 2021]
• In an HIV-negative population, CAD4TB v6 met WHO-recommended minimal accuracy for pulmonary tuberculosis triage tests [Khan et al., The Lancet Digital Health, 2020]

Diabetes

• CAD4TB offers good diagnostic accuracy as triage for TB screening among diabetes patients [Habib et al., Nature Scientific Reports, 2020]
• CAD4TB with X-ray systems allowed not only rapid and systematic triage to Xpert testing, but also found quantitatively more TB-like abnormalities in those with Diabetes Mellitus In Bangladesh [Paul et al., Science Direct, 2020]
• CAD4TB has potential as a triage tool for TB screening in people living with diabetes, thereby significantly reducing the need for microbiological examination in Indonesia [Koesoemadinata et al., IJTLD, 2018]

Prison Screening

• Inclusion of digital CXR to systematic TB screening detected additional TB cases among inmates that would otherwise have been missed, and using CAD4TB may also improve performance of screening algorithm [Kim et al., IJTLD, 2020]
• High uptake of new screening tools such as digital X-ray with CAD4TB may be particularly feasible, reliable, and highly acceptable in prison settings [Wali et al., BMC Public Health, 2019]
• CAD4TB reliably evaluates CXRs from a mostly asymptomatic prison population, with a performance comparable to local readers in Tanzania [Steiner et al., Public Health Action, 2015]